This week, Cadence Design Systems has taken a major step forward in advanced memory development by announcing the release of its integrated IP subsystem for LPDDR6. This is the industry’s first IP solution capable of operating at 14.4 gigabits per second (Gbps), significantly outperforming current LPDDR5X speeds. The announcement marks a milestone toward a new generation of chips designed for artificial intelligence, workstations, and mobile devices.
Cadence’s new IP subsystem not only supports LPDDR6 but also offers backward compatibility with LPDDR5X and the new CAMM2 package standard, a format increasingly adopted in high-performance laptop memory module designs. These features position this IP as a flexible, future-proof solution for chip design challenges.
Addressing AI Bandwidth Demands
Memory performance has become central in the era of artificial intelligence. AI accelerators, especially those processing language models and training deep neural networks, heavily rely on memory bandwidth. Until now, many solutions have used HBM (High Bandwidth Memory), but its limited capacity and high costs have driven the need for an added memory layer.
In this context, LPDDR6 appears as an ideal choice to provide increased memory capacity directly connected to the chip, complementing HBM and enabling a more efficient multilevel architecture for intensive workloads. For example, Intel showcased a design at Foundry Direct Connect 2025 featuring chiplets that integrate both HBM5 and LPDDR, targeting a hybrid architecture that would benefit directly from these innovations.
A Forward-Thinking IP
Cadence’s strategy follows a common pattern in the semiconductor industry: releasing proprietary IP packages before the hardware ecosystem is fully prepared for their adoption. Just as they already offer IP for PCI Express Gen7—when the market is only beginning to adopt PCIe Gen6—the release of this LPDDR6 IP lays the groundwork for manufacturers to start designing and integrating chiplets with this memory standard.
“These are capabilities we won’t see in commercial products for another one or two years,” explains Cliff Robinson in ServeTheHome, “but the fact that Cadence already has this IP available means that next-generation chip design and validation can start now.”
Beyond AI Applications
While the need for higher bandwidth is most urgent in AI accelerators, this technology is also poised to play a key role in other areas. Next-generation workstations, high-end mobile devices, and even advanced laptops could benefit from the power efficiency and speed advantages that LPDDR6 offers—particularly in environments where energy efficiency is critical.
Conclusion
With its new LPDDR6 IP subsystem, Cadence is ahead of the curve in meeting the demands of the next semiconductor generation, as artificial intelligence reshapes memory and processing requirements. This solution represents a significant technical milestone, enabling manufacturers to build faster, more flexible chiplets that are prepared for an era where performance and efficiency are not optional but essential.
via: serveTheHome and JEDEC